Loading...
Back to Homev1.0.0 Stable
About ClusterShot
An AI-powered zero-shot photo clustering and agentic retrieval engine built to solve high-volume event photography delivery.
Under The Hood
How ClusterShot automates ingest-to-delivery using modern machine learning and database vector indexing.
Step 01
ZIP Ingestion & Extraction
Photographers upload entire event folders inside a consolidated ZIP archive. The FastAPI server initiates an asynchronous background worker that extracts images and parses EXIF metadata (camera specs, timestamps, geotags).
Step 02
InsightFace & CLIP Embeddings
Images are analyzed by the Buffalo_L InsightFace model to detect faces and compute normalized 512D embeddings. Parallelly, OpenAI's CLIP model tags the scene (e.g. indoor, outdoor, crowd, stage).
Step 03
DBSCAN Facial Clustering
Face embeddings undergo unsupervised spatial partitioning using the DBSCAN algorithm with Cosine Distance metrics, grouping duplicate identities into singular person sub-folders automatically.
Step 04
Supabase pgvector Sync
Metadata is stored alongside face coordinates and 512D float vectors in Supabase Postgres. Original files are concurrent-uploaded to Cloudinary CDN to ensure snappy responsive loads globally.
Step 05
Vector Similarity Selfie Match
Guests upload a selfie, triggering a fast Cosine Similarity query via pgvector RPC in Supabase. The database returns the matching person ID with distance indicators within milliseconds.
Step 06
Dynamic client JSZip Export
Guests review their customized gallery, containing only photos they appear in, and export their photos instantly in one click via a client-side zipped bundle generated using JSZip.
Tech Stack & Ecosystem
A premium architectural selection ensuring low-latency processing and optimal developer ergonomics.
Frontend Architecture
- Next.js 16 (App Router, SSR)
- React 19 & TypeScript
- TailwindCSS v4
- Framer Motion (Transitions & FX)
- DnD Kit (Drag & Drop Clustering UI)
Backend & Core AI
- FastAPI (High-performance API)
- InsightFace (ONNX Face Detection)
- CLIP / Transformers (Zero-Shot Tagging)
- Scikit-Learn DBSCAN
- Gemini 2.5 Flash (Query Structuring)
Infrastructure & DB
- Supabase (Relational Database)
- pgvector Extension (Vector Similarity)
- Supabase Auth (Photographer Guard)
- Cloudinary (CDN Asset Hosting)
- Uvicorn Async Reloader Process
Meet the Developer
The mind behind ClusterShot.
Syed Muhammad Hassan
Lead AI Engineer & Full-Stack Developer•Co-Founder, AgenticEra Systems
Certified GenAI developer and 5th-semester CS student at Ziauddin University, Karachi. I build production-grade AI systems — from CNNs and RAG pipelines to multi-agent platforms. ClusterShot is my exploration of zero-shot face clustering and agentic search, built entirely end-to-end as a solo developer.
Supercharged by Gemini & Antigravity as my AI coding partners.
BUILT BY
AgenticEra Systems
Pioneering the intersection of agentic AI and real-world applications. From intelligent code assistants to multi-agent orchestration platforms — we build AI systems that augment human expertise with precision and responsibility.